Map / Core AI / Frontier Labs / TypeSafe AI

TypeSafe AI

San Francisco lab building System One models that make calibrated decisions

OrganizationNeolabs

Products by TypeSafe AI

JevTypeSafe's decision model: typed questions in, calibrated probabilities outFoundation Models

People mark and rate products, not organizations. Open a product to say you use it.

Models

Decision 1

Every model from TypeSafe AI on the map and when each version came out, grouped by what it makes. Each dot is a release.

VersionPreviewLatest version

JevLatest 1.14 · 14 Oct 20262 versions since Sep 2026

About

TypeSafe AI is a San Francisco lab, founded by former OpenAI researcher Diogo Almeida, that builds System One models: decision models that answer typed questions with calibrated probabilities instead of generating text. It trains them with its own model architecture and a method it calls Reinforcement Learning for Calibrated Decisions (RLCD). Its first model, Jev, launched in September 2026.